mozilla-foundation

/

youtube_video_similarity_model_wt

like 12

Hey,

Will you be updating this model with the second version which does not use the cross encode model? Also have you shared the dataset, I can't find the data so I can train my own model.

One issue I see is that we get a different result with:
v1 , v2
vs
v2, v1 I think we should either train on both on cross encoder or use the v2 model which would give the same result. Can you share the v2 model please.